IddSampleDriver/README.md

47 lines
3.8 KiB
Markdown
Raw Normal View History

2021-04-20 21:46:06 +00:00
License MIT and CC0 or Public Domain, whichever is least restrictive -- Use it
2021-04-08 20:30:48 +00:00
2021-04-09 21:56:10 +00:00
AS IS - NO IMPLICIT OR EXPLICIT warranty This may break your computer, it didn't break mine. It runs in User Mode which means it's less likely to cause system instability like the Blue Screen of Death.
2021-08-11 16:55:46 +00:00
Check out the latest release to download, or find other versions below:
2021-11-02 21:21:50 +00:00
https://github.com/ge9/IddSampleDriver/releases
2021-04-08 20:30:48 +00:00
2021-04-08 22:24:56 +00:00
If you want me to build on this donate eth or similar here: 0xB01b6328F8Be53c852a54432bbEe630cE0Bd559a
2021-10-06 18:43:04 +00:00
I now have a NEAR address: moopaloo.near
2021-04-08 22:24:56 +00:00
2021-04-08 22:49:00 +00:00
Thanks to https://github.com/akatrevorjay/edid-generator for the hi-res EDID.
2017-04-03 23:20:52 +00:00
# Indirect Display Driver Sample #
This is a sample driver that shows how to create a Windows Indirect Display Driver using the IddCx class extension driver.
## Background reading ##
Start at the [Indirect Display Driver Model Overview](https://msdn.microsoft.com/en-us/library/windows/hardware/mt761968(v=vs.85).aspx) on MSDN.
## Customizing the sample ##
The sample driver code is very simplistic and does nothing more than enumerate a single monitor when its device enters the D0/started power state. Throughout the code, there are `TODO` blocks with important information on implementing functionality in a production driver.
### Code structure ###
* `Direct3DDevice` class
* Contains logic for enumerating the correct render GPU from DXGI and creating a D3D device.
* Manages the lifetime of a DXGI factory and a D3D device created for the render GPU the system is using to render frames for your indirect display device's swap-chain.
* `SwapChainProcessor` class
* Processes frames for a swap-chain assigned to the monitor object on a dedicated thread.
* The sample code does nothing with the frames, but demonstrates a correct processing loop with error handling and notifying the OS of frame completion.
* `IndirectDeviceContext` class
* Processes device callbacks from IddCx.
* Manages the creation and arrival of the sample monitor.
* Handles swap-chain arrival and departure by creating a `Direct3DDevice` and handing it off to a `SwapChainProcessor`.
### First steps ###
Consider the capabilities of your device. If the device supports multiple monitors being hotplugged and removed at runtime, you may want to abstract the monitors further from the `IndirectDeviceContext` class.
The INF file included in the sample needs updating for production use. One field, `DeviceGroupId`, controls how the UMDF driver gets pooled with other UMDF drivers in the same process. Since indirect display drivers tend to be more complicated than other driver classes, it's highly recommended that you pick a unique string for this field which will cause instances of your device driver to pool in a dedicated process. This will improve system reliability in case your driver encounters a problem since other drivers will not be affected.
Ensure the device information reported to `IddCxAdapterInitAsync` is accurate. This information determines how the device is reported to the OS and what static features (like support for gamma tables) the device will have available. If some information cannot be known immediately in the `EvtDeviceD0Entry` callback, IddCx allows the driver to call `IddCxAdapterInitAsync` at any point after D0 entry, before D0 exit.
2021-04-08 20:30:21 +00:00
Careful attention should be paid to the frame processing loop. This will directly impact the performance of the user's system, so making use of the [Multimedia Class Scheduler Service](https://msdn.microsoft.com/en-us/library/windows/desktop/ms684247(v=vs.85).aspx) and DXGI's support for [GPU prioritization](https://msdn.microsoft.com/en-us/library/windows/desktop/bb174534(v=vs.85).aspx) should be considered. Any significant work should be performed outside the main processing loop, such as by queuing work in a thread pool. See `SwapChainProcessor::RunCore` for more information.